Cranial material

The sample of cranial material used in the present study originates from Liberia. It belongs to a large assemblage of chimpanzee skulls housed in the collections of the Department of Paleoanthropology in the Senckenberg Research Institute and Natural History Museum Frankfurt (Frankfurt am Main, Germany, N = 36) and the Phyletic Museum (Jena, Germany, N = 6). In total, we analyzed 100 upper and lower cheek teeth (M1: N = 47, M2: N = 41, dp4: N = 12) of 42 specimens (S1 Table). The crania originate from individuals of a free-ranging western chimpanzee (P. t. verus) population (“Liberian chimpanzees”) in the northeastern part of Nimba County (Liberia). Those individuals were killed, butchered and their skulls usually kept as trophies by locals, called Dan and Kran [46, 47], and later gathered and bought during expeditions led by the German ethnologist and medical doctor Hans Himmelheber in 1952/53 and 1955/56. The individuals originate from a restricted area, which was bounded to the east by the Cess River and to the west by the St. John River. Himmelheber reported the sex of 24 chimpanzees, which was later confirmed by Dierbach [48].

We used a second set of cranial material originating from a free-ranging chimpanzee population of the same subspecies from the Taï National Park (Ivory Coast) to address intraspecific variation in macro- and microscopic tooth wear. This cranial material is part of the osteological collection of the Taï Chimpanzee Project [49] hosted by the Departments of Human Evolution and Primatology of the Max Planck Institute for Evolutionary Anthropology (MPI-EVA Leipzig, Germany). For this population we included 95 upper and lower cheek teeth (M1: N = 49, M2: N = 26, dp4: N = 20) of 41 specimens (S1 Table). Data for the microscopic tooth wear analysis were taken from Schulz-Kornas et al. [13] and Stuhlträger et al. [14].

Habitat information

The northeastern part of Nimba County (Liberia) and the Taï National Park (Ivory Coast) comprise evergreen lowland rain forest with the latter remaining the largest one in west Africa [1, 40]. Both chimpanzee habitats are in close proximity (approximately 200 km apart, Fig 1A), and underlie similar periodic environmental changes. In the Taï forest the annual rainfall from 1950 to1975 and from 1988 to 1995 was approximately 1800 mm [1], but declined to around 1400 mm annually between 2000 and 2008 (Taï Chimpanzee Project, unpublished data). For Liberia we selected an appropriate location (N6°43’12”, W8°36’36”) and time period (1931–1960) based on historical rainfall data of the Climatic Research Unit of University of East Anglia [41], which documented an average annual rainfall of 1900 mm. In both habitats, the precipitation follows a bi-annual pattern, with two rainy periods (March—June, September—November), separated by months with a decline in rainfall. This decline is strongest from December to February. During this time the rainfall does only vary approximately from 25 mm to 48 mm per month in the Taї forest ([1]; Taï Chimpanzee Project, unpublished data), whereas in northeast Liberia the rainfall does not exceed 42 mm per month (for detailed information on the climate of the Taї forest see [1, 50]). Hence, the period from December to February is considered as a dry period [1]. Furthermore, this particular dry period is accompanied by the Harmattan, a cold desert wind from the Western Sahara, which results in dust deposition in both habitats [42–44].

Occlusal fingerprint analysis (OFA)

The OFA is a virtual method for the characterization and quantitative comparison of dental macroscopic tooth wear patterns (wear facet pattern) which result from the alteration of the occlusal primary relief in tooth crowns through dental tissue loss during the occlusal power stroke (phase I and phase II) of chewing and other dental activities [23, 27, 51]. A fully developed wear facet pattern in a hominid molar usually consists of a maximum of 13 wear facet positions [52]. Wear facets 1–4 belong to phase I buccal, wear facets 5–8 to phase I lingual, and wear facets 9–13 are attributed to phase II of the occlusal power stroke movements [28]. Here we focused on the upper and lower M1 and M2 complementary wear facets of phase II of the power stroke (Fig 1B and 1C), those wear facets should carry a surface texture induced mainly by food properties, since they are essentially worn through tooth-food-tooth contacts. During chewing, food is trapped, compressed and comminuted between antagonistic phase II facets at the end of phase I and during phase II power stroke movements [53]. To reduce the effect of age and for better comparability we only chose specimens with M1 and M2 in wear stages 3 and 4 according to Smith [54]. Consequently, and compared to the 3DST analysis, we could include only 15 specimens of the Liberian chimpanzees and 13 specimens of the Taї chimpanzees of the OFA study sample.

The OFA uses digital 3D polygonal models of crowns usually generated from high-quality dental casts [55] applying white-light 3D-surface scanning with a spatial minimum point-to-point distance in x-y-z of ~55μm (smartSCAN 3D C-5, Breuckmann GmbH). Polyworks^® 2017 (InnovMetric Inc.) 3D metrology software was used for post-processing of the computer models. The Polyworks^® Modeler module provides digital tools for a manual segmentation of each wear facets using a close polyline fitted onto the surface of the model while having the original casts under a binocular alongside for a best-practice segmentation [51]. Afterwards the polyline of all wear facets were integrated into the model surface by a re-triangulation of the model surface. Accordingly, wear facet areas were analyzed following the procedures recently described in detail by Kullmer et al. [51]. The OFA method applied here for the wear facet quantification consists of three sequential steps: determination of a reference plane through cervical margin, wear facet segmentation and mapping, and the calculation of wear facet area size and dip angle in respect to the references plane of the molar crown, while fitting a best-fit plane through wear facet surface data points.

The reference plane was calculated from a selection of data points 0.2 mm above and below a manually fitted polyline along the cervical margin (step 1). It was used to orientate the crown model in the Cartesian xy-plane [23, 27, 51, 56, 57]. The closed polyline tool and the grouping of surface data points in Polyworks^® 2017 (InnovMetric Inc.) allows the segmentation and labeling of each wear facet (step 2) on the 3D surface model. In the current study the wear facet area size value in mm² and the inclination dip angles of each phase II wear facet are calculated between the reference plane (xy-plane) and the wear facet best-fit plane (step 3). Wear facet areas of phase II facets are calculated as percentage values of the portion of area of wear facets of phase II in respect to the area percentage of all wear facets.

3D surface texture (3DST) analysis

The standard phase II wear facet for primate dietary reconstructions [29, 52] is wear facet 9 on both upper and lower molars. In this study, we selected first and second molars on either the right or left side and, where present, of the fourth deciduous premolars. The teeth were cleaned with acetone and molds were taken using a high resolution silicone imprint (Provil^® novo Light C.D.2 regular set; Type 3; Heraeus Kulzer, Dormagen, Germany), following the procedure of Schulz et al. [33, 58] and Calandra et al. [25]. We used the high-resolution confocal disc-scanning surface measuring system μsurf mobile (NanoFocus AG, Oberhausen, Germany) with a 100x long distance objective and a numerical aperture of 0.8 to acquire 3DST measurements. Three to four not overlapping scans (each 160 μm x 160 μm) of each facet with a resolution of 0.16 μm in x and y, and a step size of 0.06 μm in z were taken and averaged (mean) for further analyses. Measurements with less than 95% surface points were excluded from further analyses.

The obtained data were analyzed using 26 3DST parameters (Fig 1D, S2 Table) following the 3D areal surface texture standards ISO 25178 [59], ISO 12781 [60] as well as motif (dales and hill recognition using watershed segmentation algorithms) and furrow (surface vectorization) analyses [58] using the μsoft analysis premium software version 7.4.8076 (NanoFocus AG, Oberhausen, Germany; a derivative of Mountains^® Analysis software by Digital Surf, Besançon, France). These 3DST parameters relate to criteria such as area (Sha, Sda, mea), density (Spd, Sal, medf), height (Sq, Sp, Ssk, S5p, S5v, FLTv, FLTp, meh, madf, metf), volume (Vv, Vvc, Vmp, mev), plateau size (Smr, Smc), anisotropy (Str), texture direction (Std), complexity (Sdr), and slope (Sdq).

Statistics

For all comparisons, we analyzed upper and lower molars separately. For OFA we used both, the first and the second molar, because this method is more affected by the degree of tooth wear. For the 3DST analysis the first permanent molar was the preferred one, i.e. only if this molar was not available, we used the second molar to increase the specimen number. Deciduous fourth premolars were only used for the 3DST comparison of the two populations (dp4_Liberia vs. dp4_Taї) to test whether the possible wear differences already appear in younger individuals. All tests were run between the two populations and between sexes.

We used a circular statistic for OFA to analyze the wear facet dip angles, which we conducted in the software Oriana© (version 3.21, 1994–2010, Kovach-Computing services, Anglesey, Wales, Great Britain) [61–64]. We performed a groupwise Watson-Williams-Test between the mean angles. For the comparison of the phase II wear facet areas, we first checked whether the data are normally distributed. Therefor we used a histogram for visual exploration, as well as the Shapiro-Wilk test that affirmed a normal distribution for our data. Hence, we used the Welch Two Sample t-test in the software R 3.4.3 [65] (using the function “t.test”) to check for statistical significance. The means of wear facet dip angle and area were calculated separately for UM1, LM1, UM2 and LM2.

All the statistical analyses related to the 3DST were performed using the software R 3.4.3 [65] by using the R packages xlsx version 0.4.2 [66], doBy version 4.5.3 [67], and R.utils version 1.12.1 [68]. We used the robust Welch–Yuen heteroscedastic omnibus test (WYT) coupled with a heteroscedastic pairwise comparison test (analog to Dunnett’s T3 test). If both tests detected a significant difference (p = ≤ 0.05) a heteroscedastic rank based test (according to Cliff’s method) was used to detect significant differences between trimmed (15%) means. The trimming method excludes the highest and lowest parts of the data set and is used to compensate for non-normality [25]. For a detailed description of the statistical tests including the R-scripts see Calandra et al. [25, 69].

Additionally, we performed a Factor Analysis (FA) to compare the Liberian chimpanzees with the Taї chimpanzees from dry and rainy periods (for details see S1 Text). We selected 11 of the initial 26 3DST parameters (Sq, Sp, Vmp, Sdr, Sdq, Sal, Smc, meh, medf, metf, mea) using varimax rotation and the function “factanal”. The rotation “varimax” is recommended for a small sample size and rotates the factors in a way that each variable has a large absolute loading on only one factor, while its loadings on the other factors are close to zero [70]. The 3DST parameters were selected due to approximately normal distribution and no missing values. The Kaiser-Meyer-Olkin measure of sampling adequacy (value > 0.5), using the function “paf” of the R package rela [71], showed that the FA was justified. The FA scores of factors with Eigenvalues ≥1 were extracted to perform a one-way ANOVA with a post-hoc Tukey’s Honestly Significant Difference (HSD) test in order to explore significant variations between groups.

Inferring environmental abrasives and seasonality from tooth wear

The macroscopic tooth wear patterns of both populations are similar. However, there seems to be a trend towards steeper and smaller wear facets in Liberian chimpanzees compared to Taї chimpanzees, with an exception of the upper M2s where Liberian chimpanzee wear facets are significantly steeper and smaller. Since we exclusively used teeth with similar tooth wear stages (for details see Occlusal fingerprint analysis (OFA) in Materials and Methods) for the macroscopic tooth wear analyses, we do not think that the steeper and smaller phase II wear facets of Liberian chimpanzees can be explained by age differences among individuals of both populations. We rather suggest that steeper and smaller wear facets point to a lower rate of tissue loss in the Liberian chimpanzees, probably as a cumulative result of varying dietary properties dominated by low abrasive (a)biotic particles in a long-term perspective [19]. In contrast, more planar and larger wear facets, which are related to a higher rate of tissue loss, point toward a high abrasive diet. Previous studies on the feeding ecology of chimpanzees from regions around the Nimba Mountains, which are close to the reported distribution area of the Liberian chimpanzees studied here, reported intense consumption of different species of ants [72–74], which are considered as large abrasive dietary particles causing tooth wear [e.g. 25, 75]. In contrast, the Taї chimpanzees are well-known for their use of a hammer (e.g. branches or stones) and an anvil (e.g. root, stone) to access the soft endosperm (low abrasive) of nuts without chewing on the harder nut shells (for details see Supplement 5 in [13]). These particular and detailed dietary behaviors are not reflected in our macroscopic tooth wear results. However, it needs to be considered that first our sample size is quite small for the OFA data, and second the steeper and smaller wear facets in Liberian chimpanzees compared to Taї chimpanzees are statistically non-significant (with an exception of the upper M2) and therefore can only be treated as a trend. Hence, at this stage, it is too early to draw definite conclusions from the OFA, and it needs to be investigated further.

In contrast, the microscopic tooth wear results support the findings of previous studies that the Liberian chimpanzees, as a whole population (both sexes), presumably have fed more on insects (high density of microscopically small hill- and dale-like features) when compared to the Taї population (both sexes), who focused more on the consumption of nut endosperm. The Liberian chimpanzees in our sample show microscopically smoother wear facet surfaces with a higher density of narrow features than the Taї chimpanzees (pooled data set including dry and rainy period). A high density of microscopically small hill- and dale-like features (high values for Spd, small values for Sha, Sda) is indicative for chewing on large and hard dietary items, such as sclerotized insect exoskeletons or seeds [25]. Large hard food items fracture tooth enamel more heavily, and therefore microscopic hill- and dale-like features tend to be smaller since it is unlikely that large ones persist [25].

In addition to the inferences about the feeding preferences of Liberian chimpanzees, we found that the Liberian chimpanzees, with their many small and narrow microscopic wear features, cluster in the Factor Analysis (FA) with the Taї chimpanzees that died during dry periods (see Fig 3) (for details on the 3DST results of the Taї chimpanzees see [13]). Some of the former even have higher values on factor 2 than any of the Taї chimpanzees, implying that these Liberian individuals have the smallest and narrowest microscopic wear features of the whole data set. The habitats of both chimpanzee populations are characterized by a yearly occurring major dry period, approximately lasting from December until February [1, 41]. This dry period is accompanied by a dust-laden trade wind, the Harmattan, blowing from the Sahara Desert over West Africa into the Gulf of Guinea [43]. In fact, we know that the Harmattan reaches the Taї National Park for up to two weeks in this particular dry period, and that the forest is covered with dust particles until they are washed down by the rain at the beginning of the rainy period [1]. Despite the fact that the Liberian chimpanzees cluster with the majority of the Taї chimpanzees from Harmattan dry periods while differing from most of the Taї chimpanzees from rainy periods, both Taї chimpanzees from dry and rainy periods show some overlap in the FA, which may be accounted for as follows: Firstly, some of the Taї chimpanzee specimens died at the very beginning or end of a dry and/or rainy period. Even though the 3DST parameters reveal short-term tooth wear signals, the accumulation of new wear signals may take several days or a few weeks depending on the dietary particles [34–36]. For example, an individual that died at the beginning of the rainy period may still show microscopic wear features related to the prior period, the Harmattan dry period. Secondly, the overlap of Taї chimpanzees from dry and rainy periods might be explained by the fact that the exact arrival of the Harmattan wind in the Taї forest, which we assume as one of the major causes of the microscopic wear signals, might have varied between years (e.g. few days to one week earlier or later). Due to the lack of information of the exact arrival and duration of the Harmattan wind in each year, we used all months of the dry period (December to February) [1] and assigned them to the Harmattan dry period in our analyses.

Despite the overlap of the Taї chimpanzees from the Harmattan dry period with those from the rainy period, the former show more complex 3DSTs with a higher density of furrows compared to those Taï chimpanzees that died during the rainy period. The same applies for the Liberian chimpanzees. This is in accordance with previous studies that showed that the ingestion of abrasive components (e.g. dust) creates complex microscopic tooth wear signals with many and fine furrows and dales [13, 76]. Additionally, tooth wear and fecal particle sizes in Taї chimpanzees showed that during the dry period dust particles are consumed with the food and that their presence is reflected by a specific 3DST pattern on their molars [13]. Based on the similarities in the microscopic tooth wear of the Liberian chimpanzees and Taї chimpanzees that died during Harmattan dry periods as well as the fact that the Harmattan wind has also been present in Liberia at least since the early 20th century [77] it is highly conceivable that the Liberian chimpanzees died during such a Harmattan dry period. The chimpanzees in northeastern Liberia were reported to be under intense human predation [78]. Historical accounts from the 1950s indicate that the Liberian specimens used in this study were poached and butchered by local tribes of farmers, called Dan and Kran, and the skulls were kept as trophies [46, 47]. Even nowadays, hunting for bush meat in Nimba County (Liberia) is still a common activity by local groups; 91% of the people hunt year-round, in dry and rainy periods, but 80% of those people reported to increase their hunting activity intensively during dry periods [79]. This, in turn, is in line with our interpretation that the chimpanzees studied here most likely perished during an annual dry period, which might have accompanied by the dust-laden Harmattan trade wind. However, additional historical or cultural evidence from that time period is needed to reinforce our suggestion.

Sex-specific feeding ecology in Liberian chimpanzees

It has been widely known that chimpanzees show sex-specific feeding preferences [e.g. 1, 16, 80–82]. The macroscopic tooth wear results suggest that female chimpanzees tend to engage in a different chewing behavior than males. This is more obvious in female Taї chimpanzees where more planar and less inclined phase II wear facet surfaces are indicative of vertical chewing movements. Such chewing kinematics are thought to be required when dietary items need to be crushed (e.g. hard exoskeletons of insects, seeds) [83]. This tooth wear pattern is more pronounced on the upper second molars of Taї chimpanzees where females have significantly smaller phase II dip angles than Taї males. However, the trend for less inclined phase II wear facets in Taї females compared to males is also seen on the other molars (see Table 1). Large and less inclined phase II wear facets of Taї females indicate a high degree of repetitive vertical crushing combined with horizontal grinding activity. This could be related to an intense comminution of items like insects with hard exoskeletons and seeds embedded in fruit pulp (< 1.2 cm), on a regular basis [13].

Recently, Schulz-Kornas et al. [13] showed for the Taї chimpanzee population that female/male-specific dietary differences are reflected in the 3DST pattern. It was found that both sexes within the Taї population consume mainly fruit pulp without any hard seeds, but they differ in their feeding time on other dietary items. Females who consume more fruit pulp including large amounts of small seeds, leaves and insects have microscopically flatter and less voluminous wear facet surfaces with many micrometer-sized plateau-like features [13]. In contrast, males who feed more extensively on large seeds, vertebrate meat, and in contrast to other observations [1, 15], on a larger amount on nut endosperms, have wear facet surfaces dominated by larger dales and less plateaus [13]. However, for the comparison with our data it needs to be considered that the dietary observations of Taї male chimpanzees arise mainly from the Harmattan dry period, which may mean that the dietary and dust signal in their 3DST is mixed [13].

Our results show that Liberian females and males differ to some extent in their 3DSTs, but in a different way, than has been described for Taї females and males. Liberian females have significantly more microscopic peaks and smaller hill and dale areas as well as slightly rougher wear facet surfaces than the Liberian males, although both Liberian females and males have more peaks than Taї chimpanzees of both sexes (see Fig 4D, S6 Table). Since chewing on large and hard dietary items such as sclerotized insect exoskeletons can produce rougher wear facet surfaces and many of the microscopically small hill-like features [25], we conclude that Liberian females consumed more insects than Liberian males. This is in agreement with Smith et al. [80] who analyzed carbon signatures of the Ganta chimpanzees from northern Nimba County (Liberia). This analysis suggested a higher insect consumption in female chimpanzees and a higher meat consumption in male chimpanzees. Although the Taї females are also known to consume more insects than males [13]; the former may have consumed less insects and more softer dietary items (e.g. nut endosperm, fruit pulp) when compared to the Liberian females. Therefore, the 3DST results suggest that both populations have different feeding ecologies, at least on a short-term basis, and that this is mainly driven by the female chimpanzees of each group.